Raimo P. Hämäläinen Systems Analysis Laboratory

Aalto University, School of Science

Co-authors: Jukka Luoma and Esa Saarinen

ON THE NEED FOR BEHAVIORAL OPERATIONS

RESEARCH

Behavioral Operations Research

The study of behavioral aspects related to the use of

operations research methods in modeling, problem solving and

decision support

Behavioral research

How people behave in different settings?What are the consequences of humans being

involved?Research methods: experimental and qualitative

What is the human impact on the OR process?

“Scientific methods to improve the effectiveness of operations and systems to make better decisions”

Scientific methods:Modeling, data analysis, optimization etc.

Operations Research The Science of Better

The pioneers West Churchman and Russel Achoff:

OR is not mathematics only

What is essential in our profession?

Goal to help people in problem solving

but

Have we omitted the people,the problem owners and the OR experts,

from the analysis?

Methods and problem solving

Theory and algorithms are free of behavioral effects

but

as soon as we use them in real life problem solving behavioral effect will be present.

Model validity discussed a lot in early ORThere exists one ideal model and a good OR

specialist needs to find it.

Hidden assumption:A valid model automatically produces a valid

process and bias free objective results

Model validity The lure of objectivity

Best practices in OR Acknowledgement of

subjectivityFocus on the OR process

Based on successful of case studies First steps towards behavioral OR

So far, no behavioral researchHow do the best practices compare against

each other? Can different processes lead to different outcomes? What are the

benefits to the client?

Soft OR and Systems Thinking• Criticized OR for being too narrowly concerned

with mathematical models only• New qualitative methods for framing and

structuring• Attention to the sociology and philosophy of

modeling • Has remained mainly methodology and tool

focused with limited behavioral research

Some areas of OR have a tradition in behavioral studies

Decision and Risk Analysis• Subjectivity is explicitly taken into account• Value and utility functions to describe

preferences• Risk attitudes seeking/averse• Multicriteria evaluation of alternatives with

subjective weighting• Research on biases and

risk perceptions

Operations Management

• Studies how people act in complex decision settings

• Judgemental forecasting• Behavioural operations conference series started

in 2006• The Bullwhip effect in Supply chains - Beer game

Factory Distributor Wholesaler RetailerDelayDelayDelayDelay

Interest in behavioral issues emerges when the basic theoretical core of a

discipline has matured

Behavioral finance and economics

• What is the actual behavior of agents in economic decision making?

• How do people make personal investment decisions?

• Active research area acknowledged also by theoretical economists

• Nobel price 2002 in economics to Vernon Smith together with Daniel Kahneman

Embracing the behavioral perspective in economics

helps:“in generating theoretical insights, making better predictions, and suggesting better

policy” (Colin Camerer et al., 2004)

If this is true for economics it surely applies to OR as well

Judgement and Decision making

From: Kahneman and Tversky

• Decision theory is not enough to explain human choices

• Axioms of rationality not followed• Bounded rationality (Herbert Simon)• Prospect theory: gains and

losses seen differently (Daniel Kahneman and Amos Tversky)

• Cognitive biases• Heuristics (Gerd Gigerenzer)

From behavioral to neural• Emotions are needed in decision making• Somatic marker hypothesis (Antonio Damasio)• Brain imaging research on decision making –

neuroeconomics• How do we evaluate risks - What brain areas are

activated in risk decisions

Experimental Game Theory• How do people interact?• Ultimatum game

• The receiver should accept 1 €, 50% reject offers 20 €

• Strong tendency towards co-operative behaviour• Typically fair offers near 50 euros• Research on reciprocity and fairness• Practical implications on auctions?

Offer x €

100-x € if accept x € 0 € if reject 0 €

Split 100 €

OR is a mature discipline

We are ready to start the behavioral era!

It is natural to pay attention to howhuman behavior moderates the OR

process

OR process creates a system • Formed by the interaction of the client and the

OR analyst – usually a team• The client and the analyst are subject to

behavioral effects• The OR analyst needs to observe and understand

this system to improve its performance• A key to good practice• Use Systems Intelligence i.e. your ability to

successfully and engage with systems (Saarinen and Hämäläinen, 2004)

Social group processes in OR facilitation

• Groupthink – overconfidence (Irving Janis)• Strategic behavior by analyst

and stakeholders • Hidden agendas in modeling:

omission of factors and adverse selection of data

• Gender and cultural effects• Facilitator styles, personality etc.

This is the right model

Yes

Yes

YesYes

YesYes

Problem solving processes• What is the main intended result - learning or

optimizing?• What are the criteria used -optimizing or

satisficing?• How to facilitate when rationality cannot be

enforced?• Human behavior can seem irrational –

intransitive preferences, bounded rationality and path dependence

Research challenge

Comparative experimental research on problem solving and structuring is very difficult

Real problems can seldom be approached repeatedly

with the real decision makersExperiments with students a good first step

OR models of people behavior• People in the loop models – pilots, operators etc.• People behavior in service systems: queuing and

waiting for service • Crowd behavior in emergency situations –

Evacuation in fires, festivals

(From: Ehtamo et al)

OR models of people behavior• People in the loop models – pilots, operators etc.• People behavior in service systems: queuing and

waiting for service • Crowd behavior in emergency situations –

Evacuation in fires, festivals

(From: Ehtamo et al)

We are subject to cognitive biases

• Appeal to Authority: we tend to thoughtlessly obey those (modeling traditions) we regard as being in positions of authority

• Beauty Effect: we attribute qualities to people (models) based on their appearance

• Cognitive Dissonance: the effect of simultaneously trying to believe in two incompatible things (model/real world) at the same time

• Commitment Bias: once we are publicly committed ourselves to a position (model) we find it difficult to retreat

• Confirmation Bias: we interpret evidence to support our prior beliefs (models)

• Fundamental Attribution Error: we attribute success to our own skill (model) and failure to everyone else's skill (rivaling models)

• Inter-group Bias: we evaluate people within our own group (modelling tradition) more favorably than those outside of it

• Loss Aversion: we do stupid things to avoid realizing a loss (acknowledging failure of our modelling)

• Man With A Hammer Syndrome: some people have a single tool (model) and see every problem as a nail

• Overconfidence: we're way too confident in our abilities (models)

• Priming: exposure to some event (modelling approach) changes our response to a later event (problem needing another model)

• Representative Heuristic: we compare the under consideration (modelling approach) to whatever we happen to bring to mind

Behavioral studies in OR aim to find ways to reveal and avoid cognitive

biases in the OR process

Framing• Increasingly important when moving from

optimization to solving people related problems• Behavioral elements are strong• Definition of system boundaries and stakeholders• Stakeholders have different perspectives and

mental models• Creating a common language• A key step in many environmental problems

Model building• Usefulness of simple versus complex models• How to build models to maximize learning• Anchoring effect in selecting model scale and

reference point• Are prospect theory related phenomena relevant

when choosing the sign (increasing/decreasing) of variables

Communication with and about models

• Visual representation of system models are essential in communication

• Effects of graphs and scales used • What is the effect of educational and cultural

backgrounds of the problem owners• What can we learn from statistics?• Is software development based on behavioral

studies?

Effect of Graphical Interfaces and

Example:Simulation

Mathematica System Modeler

Vensim

True

Matlab Simulink

Behavioral research topics in OR

Teaching of OR• Balance between methods and people skills• Should every OR student learn behavioral issues?• How to teach best practices?• Developing facilitation and systems intelligence

skills• Role of software

Ethics and OR

• Ethical OR takes behavioral challenges seriously• OR is used in the most important problems of

mankind – climate models and policies• Unintentional biases in model use• Are we really solving the problem or selling our

model?• How to improve self leadership skills in OR

practice

Non-expert use of OR methods

• Modelling is a tool used in many fields• Easy OR software invites non-experts• What is the result?• What are the typical pitfalls and risks?• Who should supervise the use of OR models?• Is quick learning of the OR process possible?• Collaboration between experts and non-experts

ExampleBehavioral studies in system

dynamics

Understanding dynamics in climate change is important in modern world (John Sterman, MIT)

Why don’t well-educated adults understand

accumulation? A challenge to researchers, educators

and citizensCronin, Gonzalez, Sterman (2009)• Accumulation refers to the growth of a stock

variable when the inflow exceeds the rate of outflow

• Carbon dioxide in the atmosphere, Balance of bank accounts, Milk in the refrigerator etc.

• Experiments with the Department store task with MIT students

People entering and leaving the department store

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

During which minute did the most people enter the store?

During which minute did the most people enter the store?

96% correct answers

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

During which minute were the most people in the store?

During which minute were the most people in the store?

44% correct

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

During which minute were the fewest people in the

store? 31% correct

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

40

Minute

Peop

le /

min

ute

entering leaving

Wrong CorrectWrong

Easy to adopt a misleading starting frame

• General stock and flow system – try the general procedure and integrate the difference between the inflow and the outflow

• The department store task is a simple special case

• Computation is not required • Observe the fact that the inflow and outflow

curves intersect only once• The correct answer is obvious

Behavioural problems

• False cues which mislead the participants• Questions do not address accumulation directly• Shapes of the curves trigger inappropriate

heuristics• Availability heuristic: maximum, inflow and

outflow stand out• “Cannot be determined,” box primes to think the

task is very difficult

Re-examining the experimentAalto University students in Finland

I. Repetition of MIT procedure • Similar results

II. Revised questionnaire • Smoother curves to reduce the impact of

availability heuristic• Added questions asking about the accumulation

phenomenon directly

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 300

5

10

15

20

25

30

35

Minute

Peop

le /

min

ute

Revised smoother curves

leavingentering

Almost all of the participants were able to understand

accumulation “During which minute were the most people in the store?” (88-90% correct – originally 44%)“During which minute were there the fewest people in the store?” (72 - 76% correct – originally 31%)

People’s poor performance in the department store task does not reflect the existence of a new cognitive bias as suggested by Cronin et al.

Lesson learnt

Even the simple accumulation phenomenon can be misunderstood in the presence of distacting

triggers of biases

Extreme care needed when communicating about systems and models

SummaryBehavioral aspects influence the OR process

Framing, biases, communication, learning, group processes

The practice of OR can be improved by behavioral research

Using the term Behavioral OR will stimulate research

Behavioral OR needs to be recognized as an integral part of OR

Behavioral OR could take a leading role in advancing the responsible use of models in policy

issuesA mature field like OR becomes stronger with

behavioral research

Developing practitioner skills with a behavioral lens will keep OR alive and interesting

for our customers and the society at large

Thank you!

References and linksPresentation based on paper: R.P. Hämäläinen, J. Luoma and E. Saarinen: On the Importance of Behavioral Operational Research: The Case of Understanding and Communicating about Dynamic System, European Journal of Operational Research 2013, Vol. 228, Issue 3, pp. 623-634.

References:R.L. Ackoff: Some unsolved problems in problem solving. Operational Research Quarterly, 13:1-11, 1962C.F. Camerer and G. Loewenstein: Behavioral economics: Past, Present Future in: C.F. Camerer, G. Loewenstein, M. Rabin: Advances in Behavioral Economics, Princeton University Press, pp. 3-51. 2004. C.W. Churchman: Operations research as a profession. Management Science, 1970, 17(2), B37-B-53.M.A. Cronin, C. Gonzalez and J.D. Sterman: Why don’t well-educated adults understand accumulation? A challenge to researchers, educators, and citizens. Organizational Behavior and Human Decision Processes, 2009, 108(1), 116-130.A.R. Damasio: Descartes' Error: Emotion, Reason, and the Human Brain, London, Vintage,1994.H. Ehtamo, S. Heliövaara, T. Korhonen and S. Hostikka: Game Theoretic Best-Response Dynamics for Evacuees' Exit Selection Advances in Complex Systems, 2010, 13(1), 113-134.G. Gigerenzer, P.M. Todd and the ABC Group: Simple heuristics that make us smart, New York: Oxford University Press, 1999.R.P. Hämäläinen and E. Saarinen: Systems intelligence - the way forward? A note on Ackoff’s “Why few organizations adopt systems thinking.” Systems Research and Behavioral Science, 2008, 25(6), 821-825.

I. Janis: Groupthink: Psychological Studies of Policy Decisions and Fiascoes , Wadsworth, USA,1982.I. P. Levin, S.L. Schneider and G.J. Gaeth: All Frames Are Not Created Equal: A Typology and Critical Analysis of Framing Effects,. Organizational Behavior and Human Decision Processes, 1998, 76(2), 149-188. J. Luoma, R.P. Hämäläinen and E. Saarinen: Acting with systems intelligence: integrating complex responsive processes with the systems perspective. Journal of the Operational Research Society, 2010, 62(1), 3-11.E. Saarinen and R.P. Hämäläinen: Systems Intelligence: Connecting Engineering Thinking with Human Sensitivity. Systems Intelligence: Discovering a Hidden Competence in Human Action and Organizational Life, Systems Analysis Laboratory Research Reports. Helsinki University of Technology, 2004.H. Simon: Models of Bounded Rationality, Vol. 1. MIT Press, 502 pp,1997. J.D. Sterman: Modeling Managerial Behavior: Misperceptions of Feedback in a Dynamic Decision Making Experiment. Management Science, 1989, 35(3), 321-339. J.D. Sterman: Economics: Risk Communication on Climate: Mental Models and Mass Balance. Science, 2008, 322(5901), 532-533. D. von Winterfeldt and W. Edwards: Decision analysis and behavioral research (Vol. 1),1986, Cambridge University Press.

Systems Intelligence Research Groupwww.systemsintelligence.tkk.fi/